CELL OUTAGE DETECTION IN LTE-A CELLULAR SYSTEMS USING NEURAL NETWORKS

Authors : Hasan Tahsin OĞUZ, Aykut KALAYCIOĞLU

Pages : 31-40

View : 15 | Download : 9

Publication Date : 2018-07-31

Article Type : Research Paper

Abstract :Self-organizing networks insert ignore into journalissuearticles values(SONs); are considered as one of the key features for automation of network management in new generation of mobile communications. The upcoming fifth generation insert ignore into journalissuearticles values(5G); mobile networks are likely to offer new challenges for SON solutions. In SON structure, self-healing is an outstanding task which comes along with Cell Outage Detection insert ignore into journalissuearticles values(COD); and Cell Outage Compensation insert ignore into journalissuearticles values(COC);. This study investigates the detection of cell outages by means of the metrics generated in the User Equipment insert ignore into journalissuearticles values(UE); with the help of pattern recognition methods such as Neural Networks, Logistic Regression and k-Means algorithms. Based on the metrics like Signal to Interference Noise Ratio insert ignore into journalissuearticles values(SINR);, Reference Signal Received Quality insert ignore into journalissuearticles values(RSRQ);, and Channel Quality Indicator insert ignore into journalissuearticles values(CQI);, large amount of data is processed with supervised and unsupervised algorithms for the purpose of classifying outages and possible degradations. Our results suggest that in 79.74% of the simulation cases, Neural Network structure was able to identify the correct state of the cells whether it is outage or not with a true positive rate of 87.61% and a true negative rate of 71.87% whereas Logistic Regression gave a success rate of 78.73%, true positive rate of 88.15%, and true negative rate of 69.3%. As a future work, more sophisticated state-of-the-art deep learning mechanisms can lead us to much more successful results in cell outage detection.    
Keywords : Cell outage detection, neural networks, 5G, self organizing networks